The Changing Risks of Exposure to Gun Violence in Chicago

Charles C. Lanfear

Robert J. Sampson

David S. Kirk

Rebecca Bucci

Homicide Rates in Chicago

Enormous variation in homicide; early 1990s peak was double the period 2005-2015

Resurgence since 2017

Even during low periods, this is ten or more times the homicide rate of major European cities

Percentage of homicides using a gun has increased over time; more gun homicides now than in crime peak of early 1990s.

Questions

 

To what extent were individuals growing up in Chicago in this period exposed to gun violence?

 

And how does this exposure differ…

… by race and sex?

… by neighbourhood context?

… by cohort?

We’ll be focusing mostly on race here today. Next session will discuss cohorts.

The

Project on Human Development in Chicago Neighborhoods

• 6200 children in 7 birth cohorts
• 3 interviews from 1995–2002
• Community surveys in 1995 and 2002

PHDCN+

• Representative sample of 4 cohorts
• 1057 interviewed in 2012
• 682 followed-up in 2021

Timeline

PHDCN+ covers much longer time period than original PHDCN: Age 0 to 40, 25 years within each cohort

Well past peak criminal involvement and exposure ages

Contexts of Violence

Align by age instead of year—see age-specific violence context

Age 13 to 25 highlighted as peak risk period

4 different patterns: High decline, medium decline, stable low, rising

Note that many other changes were occurring than just violence—incarceration, arrests, the economy. I’ll discuss those more in the talk in the next session.

Lifecourse Exposures to Gun Violence

Measures

Has the respondent…

• ever seen someone else get shot?
  • If so, at what age? (only wave 2)
• ever been shot?
  • If so, at what age? (waves 2 & 5)

Problem: Interval censoring

Method: Non-parametric MLE (Turnbull 1976)

Interval censoring means we don’t know exactly when they saw someone shot if they

Seen shot: race/ethnicity

Long plateaus are due to interval censoring: The upticks immediately afterward occurred somewhere in that period.

Note that seeing people shot typically occurs before age 20

60% of black and hispanic respondents seen someone shot by their 40s. Under 30% for whites.

While exposure lower for white respondents, it still ticks up in later ages, which are all the recent period of elevated violence

Been shot: race/ethnicity

Being shot occurs later on average—mostly after 18 or so

No white respondents were shot after age 22, while black and Hispanic respondents saw exposures in recent period

Seen shot: sex

Men are somewhat more likely to see a shooting, but general pattern is similar across age

Been shot: sex

Sex divergence is very large for being shot

Risk and Protective Factors

Additional Measures

• Cohort
• Immigrant generation
• Childhood neighbourhood context
  • Collective efficacy
  • Disadvantage
  • Homicide rate

Method: Semi-parametric proportional hazards MLE (Anderson-Bergman 2017)

Estimates

Been shot is rare, so statistical power is low

Most patterns similar across outcomes, except cohort 0

Here the cohorts are proxying for any historical differences

This includes different overall city levels of violence

Very modest relationship between exposure and neighborhood measures; trumped by individual factors

Takeaways

• Being shot occurs later in life than seeing shootings

• Only Black and Hispanic respondents were shot in adulthood

• Seeing shootings much less common for white respondents

• Mainly males are shot, but female respondents witness nearly as many shootings

Appendix

References

Sampson, Kirk, & Bucci. 2022. “Cohort Profile: Project on Human Development in Chicago Neighborhoods and Its Additions (PHDCN+).” Journal of Developmental and Life-Course Criminology 8.

Turnbull. 1976. “The empirical distribution function with arbitrarily grouped, censored and truncated data.” Journal of the Royal Statistical Society: Series B (Methodological) 38(3):290-295.

Anderson-Bergman. 2017. “icenReg: Regression models for interval censored data in R.” Journal of Statistical Software 81(12):1–23.